Using genetic risk scores in type 1 diabetes
A new review in Diabetologia looks at how genetic risk scores are already being applied, both at the pre-clinical stages of type 1 diabetes and to more accurate classification of type 1, type 2 and other forms of diabetes. This exciting technology could lead to more accurate diagnosis and tailored care. Dr Susan Aldridge reports.
Type 1 diabetes develops in stages, with genetic risk and environmental triggers contributing to the start of autoimmunity, which leads to the presence of islet-specific autoantibodies. Stage 1 is marked by the presence of these autoantibodies, stage 2 involves progression to dysglycaemia, while stage 3 is clinical diabetes.
More than 70 genetic regions have now been associated with type 1 diabetes, with particularly strong associations in the human leukocyte antigen (HLA) Class II regions. In the past, an individual’s genetic risk was assessed by family history, HLA typing or genotyping of other diabetes-associated loci. However, the increased ease and lower cost of acquiring genetic data has led to the development of scores, which combine risk variants from associated loci into a single number.
The polygenic risk score (PRS) includes the risk of many possibly correlated variants from across the genome, even if they do not reach genome-wide significance. The genetic risk score (GRS) estimates the cumulative contribution of a smaller subset of genetic variants that do reach genome-wide significance. Richard Oram of the University of Exeter and colleagues elsewhere describe the genetic architecture of type 1 diabetes, focusing on the development of GRSs. They look at the application of the risk scores in clinical studies and the opportunities and challenges involved in their translation into mainstream diabetes care.
Heritability of type 1 diabetes
Twin and family studies provide evidence for a substantial heritable component in type 1 diabetes, which declines with age at diagnosis. Concordance rates for monozygotic and dizygotic twins are more than 50% and around 8%, respectively. Sibling concordance rates range from 6-10%, while there is a risk of 6-9% for offspring of an affected father and 1-4% for those of an affected mother.
Research has attributed a large amount of type 1 diabetes heritability to variation in the class II HLA genes located within the major histocompatibility complex (MHC) region on chromosome 6. Some specific haplotypes, (combinations of alleles at multiple loci on the same chromosome) that confer the highest risk of type 1 diabetes have been identified.
Meanwhile, other studies have identified more than 70 non-HLA type 1 diabetes risk loci. The greater than 50% heritable risk explained by the variants in the HLA region is similar to that seen in other autoimmune diseases, but differs from that seen in other complex diseases, such as type 2 diabetes, where heritability is more equally distributed across all chromosomes.
Identifying variants with small effect sizes, which can combine to have an additive effect on disease risk, is necessary to uncover the full range of genetic risk in order to fully understand the pathogenesis of the disease. This is the rationale behind the development of the genetic risk score.
Genetic risk scores for type 1 diabetes
A number of GRSs have already been developed, making use of both HLA haplotypes and non-HLA variants. For instance, the author and his team have put together a 30 single nucleotide polymorphism (SNP) GRS known as GRS1. Using SNPs as ‘tags’ for key HLA alleles has the advantage of reducing the cost of genetic screening. GRS1 proved highly effective in discriminating type 1 from type 2 diabetes. It was then combined with another GRS that included 12 non-HLA loci and applied in The Environmental Determinants of Diabetes in the Young (TEDDY) study, which is looking at type 1 diabetes progression in individuals with high HLA risk. This combined GRS only needed cheap custom SNP assays, opening the door to population-wide genetic screening.
Meanwhile, GRS2 is another GRS tool, which includes some extra variants and tags, intended to capture further HLA gene contributions to genetic risk. It has been applied in the Type 1 Diabetes Genetics Consortium where its ability to identify cases was close to 100% accurate. The highest discrimination was found at younger ages, emphasising the importance of capturing HLA-based risk and the stronger contribution of genetics in very young children.
The GRS models described above were derived from individuals of European ancestry and may not have the same predictive power among those of different ancestries. For example, these early GRSs have worked well in people from Europe, South Asia and Iran and in Hispanic populations, but were less discriminative in African-American populations. Accordingly, an African ancestry type 1 diabetes GRS has been developed and has now been validated in several cohorts. There is also research developing GRSs more specific to East Asian and other populations, opening up the prospect of eventually making screening for type 1 genetic risk available and relevant world wide.
GRS for classification of diabetes
The increasing prevalence of obesity has led to more people being diagnosed with type 2 diabetes at a younger age, while levels of obesity among those who have type 1 diabetes is also on the increase. Moreover, type 1 diabetes can present at any age and, despite popular belief, onset during adulthood is actually more frequent than during childhood. All of this makes distinguishing between the two types of diabetes challenging, but using a GRS in addition to clinical features and antibody status could make for a more accurate diagnosis.
GRS1 can distinguish between type 1 diabetes and type 2 diabetes and between type 1 diabetes and monogenic diabetes (MODY). And, when combined with clinical features and biomarkers, it provided the best discrimination of progression to severe insulin deficiency. This approach has been validated among those with adult-onset type 1 diabetes and GADA antibody-positive type 2 diabetes. Genetic risk scores can also distinguish MODY from type 1 diabetes (as mentioned above), as a high GRS score is indicative of type 1 and a lower score suggests MODY. So adding a GRS to antibody and C-peptide testing can help identify those with monogenic diabetes.
Recent research using GRS tools is also providing new insights into early-onset diabetes, helping distinguish between early-onset type 1 diabetes and monogenic diabetes. It is also being applied in adult-onset type 1 diabetes, particularly in the identification of latent autoimmune diabetes of adults (LADA), which has genetic signals associated with both type 1 and type 2 diabetes. Here, a type 1 diabetes GRS was found to predict LADA more accurately than a type 2 diabetes GRS, showing that LADA is more similar to type 1 than type 2 diabetes.
Looking to the future
Now that disease-modifying therapy in the form of teplizumab is available for those in the early stages of type 1 diabetes, population screening to identify those at risk has never been more important. Including a type 1 diabetes GRS as part of a newborn screen could help those who would benefit from more intense monitoring going forward. Simulation studies with GRS2 showed that over 77% of future type 1 diabetes cases could be identified in 10% of the general population, showing the value of this approach. Another study has shown that children in the TEDDY study with a specific HLA haplotype and a GRS of greater than 14.4 had an 11% risk of developing multiple autoantibodies by the age of six, while those who had a lower GRS had a corresponding lower risk of 4.1%.
Accurate prediction models generally include multiple, time-dependent clinical biomarkers, such as autoantibodies in type 1 diabetes. Combining multiple predictive factors, such as family history, autoantibody status and GRSs may improve the prediction of type 1 diabetes through a combined risk score (CRS). A study has shown that using a CRS in nearly 8000 high-risk children aged two years significantly enhanced the prediction of type 1 diabetes, compared with the use of autoantibody status alone. Meanwhile, a 30 SNP GRS combined with age, autoantibody status and Diabetes Prevention Trial-Type 1 (DPT-1) Risk Score predicted progression to type 1 diabetes in at-risk family members, initially without diabetes, in the TrialNet Pathway to Prevention programme.
It is the pace of research in genetics that has driven the development of GRSs. SNP genotyping is becoming ever cheaper, with GRS development costs being related mainly to blood sampling and DNA extraction. Once performed, SNP genotyping information can be interpreted across an individual’s lifetime, with no need for repeat testing. The discriminative power of most type 1 diabetes GRSs is attributed to just nine SNPs, so determining type 1 diabetes risk will become ever cheaper. And with the increasing availability of direct-to-consumer testing, individuals are starting to gain access to their own genetic information. It may be that type 1 diabetes GRSs will find their way into these tests, although there are obviously risks involved with respect to direct consumer interpretation of the results.
At present, there are limited guidelines and regulations around GRSs and these will need to be put in place to ensure accuracy and appropriate application in clinical practice. There is considerable potential for identification of high-risk individuals who could benefit from follow-up care and treatment. However, there is a need for further research into how GRSs should be implemented into diabetes care.
In conclusion, advances in genetics have allowed the development of GRS models that can distinguish between type 1 diabetes and other diabetes phenotypes. This technology has significant potential for ‘test and treat’ approaches that could tailor care for individuals with type 1 diabetes.
To read this paper, go to: Luckett AM, Weedon MN, Hawkes G, Leslie RD, Oram RA, Grant SFA. Utility of genetic risk scores in type 1 diabetes. Diabetologia 13 July 2023. https://doi.org/10.1007/s00125-023-05955-y
To learn more, enrol on the EASD e-Learning course ‘Phenotypic variability’.
Any opinions expressed in this article are the responsibility of the EASD e-Learning Programme Director, Dr Eleanor D Kennedy.